Abstract
5G/B5G is envisioned to support various services with the assistance of network slices, each slice instance asks for adequate resources to provide the pre-negotiated service quality to its subscribers. Slice Admission Control (SAC) algorithm is a necessity for Slice Providers (SPs) to guarantee the QoS and QoE of each admitted request with limited resources. In that circumstance, the priority concern of services and the fairness of resource allocation arise as meaningful topics for researchers. The former originates from the innate characteristics of various services supported by 5G networks, and the latter matters because slices are instantiated on shared physical equipment. However, the two issues are mainly investigated separately in the literature or do not receive sufficient research simultaneously. In this work, we study the SAC problem in 5G/B5G networks, aiming at enhancing the fairness degree on the premise of satisfying the necessary priority requirements. We first reinterpret priority as a higher cumulative service acceptance ratio (CSAR), and adopt the uniformity of adjacent CSAR gaps to reflect the fairness. Based on these adjustments, the SAC problem is formulated as a non-linear and non-convex multi-objective optimization. Thus, we propose a heuristic algorithm called Prioritized Slice Admission Control Considering Fairness (PSACCF) to solve it. It introduces the resource efficiency of services to amend priority violations, then promotes fairness by setting the target CSARs for each service type and pushing their actual CSARs toward. Numerous simulations are carried out to compare the performance of PSACCF with two existing algorithms, termed MHPF and AHPF. Results show that our algorithm can achieve a nearly identical priority indicator to the comparisons, as well as at least a 33.6% improvement in fairness degree and a higher minimum average resource utilization.
Introduction
In 5G and Beyond 5G (B5G) networks, network slicing is the foundational technology that allows a single physical infrastructure to be partitioned into multiple virtual networks. Each "slice" is tailored to specific service requirements, ranging from high-bandwidth video streaming (eMBB) to ultra-reliable low-latency communications (uRLLC). As shown in the classification of applications, these services vary significantly in their performance and resource elasticity.
The core challenge for Slice Providers (SPs) is Slice Admission Control (SAC). When network resources are limited, the SP must decide which requests to admit to maximize resource utilization while respecting service priorities and ensuring fairness among different tenants.
System Model and Problem Formulation
The proposed system model considers a multi-tier architecture involving subscribers, tenants, and the Slice Provider. A critical aspect of this model is the inclusion of "impatient" subscriber behaviors: balking (refusing to join a long queue) and reneging (leaving a queue after a certain wait time). These behaviors are often ignored in traditional models but are vital for representing human-centric network usage.
Traditionally, priority and fairness are seen as conflicting objectives. High-priority services often starve low-priority ones. The authors address this by formulating SAC as a multi-objective optimization problem that seeks a balance between these two metrics, redefining fairness as the uniformity of gaps between service acceptance ratios.
The PSACCF Algorithm
The Prioritized Online Slice Admission Control Considering Fairness (PSACCF) algorithm is a two-stage heuristic designed to solve the non-linear optimization problem:
- Priority Amendment: The algorithm ensures that high-priority services maintain a higher Cumulative Service Acceptance Ratio (CSAR) than lower-priority ones. It uses "Resource Efficiency"—the derivative of the acceptance ratio relative to resource consumption—to prioritize admissions that fix priority violations.
- Fairness Enhancement: Once priorities are met, the algorithm optimizes fairness. Instead of aiming for equal acceptance across all services (which would ignore priority), it aims for uniform gaps between the CSARs of different priority levels. This "fairness in inequality" ensures that no service type is completely neglected.
Key Results
The performance of PSACCF was evaluated against two benchmarks: Maximum High Priority First (MHPF) and Aggressive High Priority First (AHPF). Key findings include:
- Fairness Improvement: PSACCF achieved at least a 33.6% improvement in the fairness degree compared to benchmarks.
- Resource Utilization: The algorithm improved minimum average resource utilization by up to 13.4% in high-load scenarios.
- Operational Efficiency: Decision times remained between 1ms and 4ms, confirming the algorithm's suitability for real-time 5G environments.
- Stability: Unlike aggressive preemption methods, PSACCF remained stable even when subscribers exhibited high rates of balking and reneging.
Conclusion
PSACCF successfully bridges the gap between priority and fairness in 5G slice admission control. By redefining fairness as the uniformity of acceptance gaps, the algorithm provides a more nuanced and effective way to manage shared infrastructure. Future work will explore multi-provider environments and the economic trade-offs between profit maximization and extreme fairness.
Summary updated